CN105872565A

CN105872565A - Image encoding device, image decoding device, image encoding method, and image decoding method

Info

Publication number: CN105872565A
Application number: CN201610467157.4A
Authority: CN
Inventors: 峯泽彰; 杉本和夫; 关口俊; 关口俊一
Original assignee: Mitsubishi Electric Corp
Current assignee: Mitsubishi Electric Corp
Priority date: 2011-01-12
Filing date: 2012-01-06
Publication date: 2016-08-17
Anticipated expiration: 2032-01-06
Also published as: WO2012096150A1; US9414073B2; MX2013008119A; RU2565038C2; CN105872566B; EP2665274A1; SG10201810141WA; EP2665274A4; US20160156931A1; CN103299637B; KR20160087925A; SG10202008690XA; US10931946B2; KR101643528B1; JP2017118571A; SG10201403553XA; US20160309147A1; KR101547041B1; TW201624420A; TW201724023A

Abstract

The invention provides an image encoding device, an image decoding device, an image encoding method, and an image decoding method. When carrying out an intra-frame prediction process to generate an intra prediction image by using an already-encoded image signal in a frame, an intra prediction part (4) refers to a table for filter selection to select a filter from one or more filters which are prepared in advance according to the states of various parameters associated with the encoding of a target block to be filtered, and carries out a filtering process on the prediction image by using the filter. As a result, prediction errors locally occurring can be reduced, and the image quality can be improved.

Description

Picture coding device and method and image decoder and method

The application is Application No. 201280005061.6 (PCT/JP2012/000061), the applying date It is on January 6th, 2012 (date of filing is on July 11st, 2013), invention entitled " motion Picture coding device, dynamic image decoding device, dynamic image encoding method and moving image are translated Code method " the divisional application of application for a patent for invention.

Technical field

The present invention is to compile to the picture coding device and image carrying out picture coding about with high efficiency Code method and with high efficiency to be coded of image decoder that image decodes and The technology of image decoding method.

Background technology

Such as, MPEG (Moving Picture Experts Group: Motion Picture Experts Group), In the international standard Video coding mode of " ITU-T is H.26x " etc. so that input video frame is divided into The block (coded object block) of rectangle, the image encoded by implementing use to this coded object block The prediction of signal processes and generates prognostic chart picture, by as this coded object block and prognostic chart picture The predictive error signal of difference in units of block, carry out orthogonal transformation, quantification treatment, carry out letter Breath compression.

Such as, at the MPEG-4AVC/H.264 (ISO/IEC as international standard mode 14496-10 │ ITU-T is H.264) in, carry out at the infra-frame prediction of encoded neighbouring pixel Manage or process (for example, referring to non-patent literature 1) at the motion compensated prediction adjacent to interframe.

In MPEG-4AVC/H.264, in the intra prediction mode of brightness, can be with block Unit selects a predictive mode from multiple predictive modes.

Figure 10 is to represent that the block size of brightness is saying of the intra prediction mode in the case of 4 × 4 pixels Bright figure.

In Fig. 10, the pixel in white circle is the block of coded object.Black circle is for prediction Pixel, is the pixel in encoded adjacent block.

In Fig. 10, as intra prediction mode, ready mode 0～pattern 8 totally 9 patterns, and Pattern 2 is by the pattern of meansigma methods prediction, utilizes neighbor average of top and left side block Value carrys out the pixel in predictive coding object block.

Pattern beyond pattern 2 is by the pattern of directional prediction.Pattern 0 is vertically oriented pre- Survey, generate prognostic chart picture by repeating the neighbor of top block in vertical direction.Such as, exist Pattern 0 is selected when being nicking style.

Pattern 1 is horizontally oriented prediction, by repeating the neighbor of left side block in the horizontal direction and Generate prognostic chart picture.Such as, pattern 1 is selected when being horizontal stripe style.

Pattern 3～pattern 8 use top or the neighbor of left side block, at the direction (arrow of regulation Head represent direction) generate interpolating pixel thus generate prognostic chart picture.

The block size of the brightness of application infra-frame prediction can be from 4 × 4 pixels, 8 × 8 pixels, 16 × 16 pictures Element selects, in the case of being 8 × 8 pixels, the same with the situation of 4 × 4 pixels specifies 9 frames Inner estimation mode.

In the case of being 16 × 16 pixels, it is stipulated that 4 intra prediction modes (meansigma methods prediction, Vertical direction prediction, horizontal direction prediction, planar prediction).

Planar prediction is will to enter the neighbor of top block and the neighbor of left side block at tilted direction Row interpolation interpolation and the pixel that generates are as the pattern of predictive value.

The intra prediction mode of travel direction prediction is because waiting at such as 45 degree and being predefined by pattern Direction generate predictive value, so the direction at the border of the object in block (edge) with prediction mould In the case of direction shown in formula is consistent, it was predicted that efficiency uprises and can reduce encoding amount.

But, the most only produce slightly between the direction at edge and the direction shown in predictive mode Deviation or i.e. arrange but edge in coded object block slightly deforms and (rocks, bends Deng), big forecast error will occur locally, it was predicted that efficiency drastically declines.

In order to prevent the decline of this predictive efficiency, in the directional prediction of 8 × 8 pixels, pass through Use the mode that encoded neighbor is implemented smoothing processing to be predicted processing, generate and put down The prognostic chart picture slided so that reduce in the slightly deviation producing prediction direction, produce slightly at edge The forecast error occurred in the case of being permitted deformation.

Non-patent literature 1:MPEG-4AVC (ISO/IEC 14496-10)/ITU-TH.264 marks Accurate

Because conventional picture coding device is configured as described above, if so generating and be smoothed Prognostic chart picture, even if then the slightly deviation or produce at edge in Occurrence forecast direction slightly deforms, It also is able to reduce the forecast error occurred.But, in non-patent literature 1, in 8 × 8 pixels Block outside do not implement smoothing processing, the block of 8 × 8 pixels the most only carries out general smoothing processing.

Even if it practice, the block of size beyond 8 × 8 pixels there is problems in that similarly Even if i.e. prognostic chart picture is similar with the pattern of encoded object image, due to edge small not Join and big forecast error occurs the most locally, produce the significantly decline of predictive efficiency.It addition, There is a problem in that: even if in the block of same size, if predictive error signal is being carried out The quantization parameter that uses during quantization, the difference such as the position of pixel, predictive mode in block, then be applicable to The process of the forecast error reducing locality is the most different, but is only prepared with general smoothing processing and nothing Method is substantially reduced forecast error.

And then, there is problems: when having carried out meansigma methods prediction, pre-due to by block Survey all as the meansigma methods of the pixel adjacent with block, so being positioned at the prediction letter of the pixel of block boundary Number easily become discontinuous signal relative to the encoded pixel of surrounding, on the other hand because figure Image signal is usually the signal that direction in space dependency is high, so easily existing because of above-mentioned discontinuity The boundary member of block produces forecast error.

Summary of the invention

The present invention proposes to solve problem as described above, its object is to obtain one The forecast error occurred locally can be reduced and improve the picture coding device of image quality, figure As code translator, method for encoding images and image decoding method.

About the picture coding device that the present invention relates to, in intraprediction unit is by using frame Encoded picture signal, implement intra-prediction process and generate prognostic chart as time, from preparing in advance The wave filter of more than 1 in, according to the various parameters involved by the coding of Filtering Processing object block State select wave filter, use this wave filter, implement the Filtering Processing for prognostic chart picture, Prognostic chart picture after Filtering Processing is exported to difference image signal generating unit.

According to the present invention, because being configured to encoded by use in frame in intraprediction unit Picture signal, implement intra-prediction process generate prognostic chart as time, from pre-prepd 1 with On wave filter in, come according to the state of the various parameters involved by the coding of Filtering Processing object block Select wave filter, use this wave filter to implement the Filtering Processing for prognostic chart picture, by Filtering Processing After prognostic chart picture export to difference image signal generating unit, send out locally so having to reduce Raw forecast error and improve the effect of image quality.

Accompanying drawing explanation

Fig. 1 is the pie graph of the dynamic image encoding device representing embodiments of the present invention 1.

Fig. 2 is the pie graph of the dynamic image decoding device representing embodiments of the present invention 1.

Fig. 3 is the process content of the dynamic image encoding device representing embodiments of the present invention 1 Flow chart.

Fig. 4 is the process content of the dynamic image decoding device representing embodiments of the present invention 1 Flow chart.

Fig. 5 is the appearance representing and maximum sized encoding block being hierarchically divided into multiple encoding block Explanatory diagram.

(a) of Fig. 6 represents the distribution of the subregion after segmentation, (b) be with 4 pitch tree graphs represent by Coding mode m (Bⁿ) distribute to be layered the explanatory diagram of the situation of the subregion after splitting.

Fig. 7 is to represent at encoding block BⁿInterior each subregion P_i ⁿIn the infra-frame prediction parameter that can select The explanatory diagram of one example of (intra prediction mode).

Fig. 8 is to represent at l_i ⁿ=m_i ⁿWhen=4, generate subregion P_i ⁿIn the predictive value of pixel time use The explanatory diagram of an example of pixel.

The explanatory diagram of one example of reference pixels configuration when Fig. 9 is to represent N=5.

Figure 10 is the infra-frame prediction mould representing the non-patent literature 1 when luminance block size is 4 × 4 pixel The explanatory diagram of formula.

Figure 11 be represent generate prognostic chart as time use frame in encoded image and filtering at The explanatory diagram of one example of the distance of reason subject pixels.

Figure 12 is to represent the explanatory diagram that the reference pixels of concrete wave filter configures.

Figure 13 is to represent to determine intra prediction mode index and each combination of the combination of partitions sizes Use the explanatory diagram of an example of the table of which wave filter.

Figure 14 is the example representing the simplification being filtered process when having carried out meansigma methods prediction Explanatory diagram.

Figure 15 is represent bit stream when wave filter selects table index be added to sequence-level title one The explanatory diagram of example.

Figure 16 is represent bit stream when wave filter selects table index be added to picture level title one The explanatory diagram of example.

Figure 17 is the example representing bit stream when wave filter selects table index be added to slice header The explanatory diagram of son.

Figure 18 is represent bit stream when wave filter selects table index be added to reference block title one The explanatory diagram of example.

Figure 19 is to represent with decision intra prediction mode index and each group of the combination of partitions sizes Close the explanatory diagram of the example different for Figure 13 of the table using which wave filter.

Figure 20 is that each combination representing the combination to intra prediction mode index and partitions sizes determines Whether saying the example of the table that the reference pixels when generating medium range forecast image is smoothed Bright figure.

Symbol description

1: coding-control portion (coding control unit)

2: block cutting part (block cutting unit)

3: switching switch (intraprediction unit, motion compensated prediction unit)

4: infra-frame prediction portion (intraprediction unit)

5: motion compensated prediction portion (motion compensated prediction unit)

6: subtraction portion (difference image signal generating unit)

7: change quantization portion (image compression unit)

8: inverse quantization inverse transformation portion

9: adder

10: infra-frame prediction memorizer

11: loop filter portion

12: motion-compensated predicted frames memorizer

13: variable length code portion (variable length coding unit)

51: variable-length decoder unit (variable-length decoding unit)

52: switching switch (intraprediction unit, motion compensated prediction unit)

53: infra-frame prediction portion (intraprediction unit)

54: motion compensated prediction portion (motion compensated prediction unit)

55: inverse quantization inverse transformation portion (difference image signal generating unit)

56: adder (decoded picture signal generating unit)

57: infra-frame prediction memorizer

58: loop filter portion

59: motion-compensated predicted frames memorizer

100: wave filter selects table index

Detailed description of the invention

Hereinafter, in order to illustrate in greater detail the present invention, embodiment is illustrated according to accompanying drawing.

Embodiment 1

Dynamic image encoding device and dynamic image decoding device are said by present embodiment 1 Bright, each two field picture of this dynamic image encoding device input video, by implementing from encoded Near pixel intra-prediction process or between contiguous frames implement motion compensated prediction process and give birth to Become prognostic chart picture, at the predictive error signal to the difference image as this prognostic chart picture and two field picture After implementing compression based on quantized orthogonal transform process, carry out variable length code and generate bit Stream, the bit stream exported from this dynamic image encoding device is translated by this dynamic image decoding device Code.

The dynamic image encoding device of present embodiment 1 is characterized in that: is adapted to video signal The change of the locality in space time direction, is divided into the region of sizes to carry out by video signal Adaptive coding between intra frame.

Usually, video signal has the complexity of signal and changes locally on space time Characteristic.Spatially see, in a certain specific frame of video, existing on high, wall etc. that The image-region that sample is relatively wide has the pattern of uniform characteristics of signals, has also been mixed at tool There is in the little image-regions such as the drawing having personage, careful texture the pattern of complex texture pattern.

Even if seeing in time, although sky, the change locality of pattern of time orientation of wall Ground is little, but the personage of motion, its profile of object are because carrying out rigidity and non-rigid motion in time, So temporal change is big.

The forecast error that coded treatment generates signal electric power by the prediction on time and space, entropy is little Signal, thus cut down the encoding amount of entirety, and if able to the parameter being used for prediction equably should For the biggest picture signal region, then can reduce the encoding amount of this parameter.

On the other hand, for spatially changing big picture signal pattern on the time, if application is same One Prediction Parameters, then because the mistake of prediction increases, so the volume of predictive error signal cannot be cut down Code amount.

Accordingly it is desirable to for spatially changing big picture signal pattern on the time, reduce prediction right The region of elephant, even if increasing the data volume of the parameter for prediction, also reduces predictive error signal Electric power, entropy.

In order to carry out being adapted to the coding of the general aspects of such video signal, in present embodiment 1 Dynamic image encoding device in, from regulation maximum lump size layering divided video signal region, Each cut zone is implemented prediction process, the coded treatment of forecast error.

The dynamic image encoding device of present embodiment 1 is video as the video signal of process object The arbitrary video signal that frame has level, the numeral sample (pixel) of vertical 2 dimensions arranges, except There is luminance signal and the YUV signal of 2 colour difference signals, RGB from the output of digital camera element Beyond the colour-video signal of the arbitrary color space of signal etc., also monochromatic image signal, red Outside line picture signal etc..

The gray scale of each pixel both can be 8, it is also possible to is 10, the gray scale of 12 etc..

But, in the following description, unless otherwise prescribed, it is assumed that the video signal of input is YUV Signal.In addition, it is assumed that 2 aberration composition U, V are the 4:2:0 of sub sampling relative to luminance components Y The signal of form.

And, the process data unit corresponding with each frame of video is referred to as " picture ", in this enforcement In mode 1, " picture " is as carrying out sequential scan (progressive scanning: progressive scan) The signal of frame of video illustrate.But, in the case of video signal is interleaving signal, " figure Sheet " can also be the field picture signal as the unit constituting frame of video.

In FIG, coding-control portion 1 implements the following processing: be decided to become at enforcement infra-frame prediction Reason (intra-prediction process) or motion compensated prediction process process time (inter prediction process) The full-size of the encoding block of unit, and determine when the layering of maximum sized encoding block is split The upper limit number of plies.

It addition, coding-control portion 1 implements the following processing: from the coding of utilizable more than 1 In pattern (intra-frame encoding mode of more than 1, the interframe encoding mode of more than 1), select suitable Coding mode in each encoding block of layering segmentation.

It addition, coding-control portion 1 implements the following processing: each to each encoding block, determine to exist The quantization parameter used during compressed differential image and transform block size, and determine implementing prediction The infra-frame prediction parameter used during process or inter prediction parameter.Quantization parameter and transform block chi Very little it is contained in coded prediction error parameter, is exported to change quantization portion 7, inverse quantization inverse transformation portion 8 and variable length code portion 13 etc..

And, coding-control portion 1 constitutes coding control unit.

Block cutting part 2 implements the following processing: when input represents the video signal of input picture, will The input picture that this video signal represents be divided into by coding-control portion 1 determined maximum sized Encoding block, and until reaching the upper limit number of plies determined by coding-control portion 1, layering segmentation This encoding block.And, block cutting part 2 constitutes block cutting unit.

Switching switch 3 implements the following processing: if the coding mode selected by coding-control portion 1 is Intra-frame encoding mode, then export infra-frame prediction portion 4 by the encoding block split with block cutting part 2, If the coding mode selected by coding-control portion 1 is interframe encoding mode, then will be by block cutting part 2 The encoding block split exports motion compensated prediction portion 5.

Infra-frame prediction portion 4 implements the following processing: when receiving by block cutting part 2 from switching switch 3 During the encoding block split, use the encoded picture signal in frame, according to from coding-control portion 1 The infra-frame prediction parameter of output, generates pre-for the intra-prediction process of this encoding block by implementing Altimetric image.

But, infra-frame prediction portion 4 is after generating above-mentioned prognostic chart picture, from pre-prepd 1 In above wave filter, generate with above-mentioned prognostic chart as identical according in dynamic image decoding device Prognostic chart picture moment point known to the state of various parameters select wave filter, use this filtering The Filtering Processing for above-mentioned prognostic chart picture implemented by device, is exported by the prognostic chart picture after Filtering Processing To subtraction portion 6 and adder 9.

Specifically, as above-mentioned each parameter according to more than at least 1 in following 4 parameters The state of parameter determines wave filter uniquely.

Parameter (1)

The block size of above-mentioned prognostic chart picture

Parameter (2)

The quantization parameter determined by coding-control portion 1

Parameter (3)

Generate prognostic chart as time use frame in encoded picture signal and Filtering Processing object The distance of pixel

Parameter (4)

The infra-frame prediction parameter determined by coding-control portion 1

And, switching switch 3 and infra-frame prediction portion 4 constitute intraprediction unit.

Motion compensated prediction portion 5 implements the following processing: as be suitable to have been split by block cutting part 2 The coding mode of encoding block, in the case of being have selected interframe encoding mode by coding-control portion 1, Use the reference image of more than 1 frame kept by motion-compensated predicted frames memorizer 12, according to from volume The inter prediction parameter of code control portion 1 output, pre-for the motion compensation of this encoding block by implementing Survey processes and generates prognostic chart picture.And, by switching switch 3 and motion compensated prediction portion 5 structure Become motion compensated prediction unit.

Subtraction portion 6 implements the following processing: by subtracting from the encoding block split by block cutting part 2 Remove the prognostic chart picture generated by infra-frame prediction portion 4 or motion compensated prediction portion 5, generate difference diagram As (=encoding block-prognostic chart picture).And, subtraction portion 6 constitutes difference image signal generating unit.

Change quantization portion 7 is comprised with the coded prediction error parameter exported from coding-control portion 1 (such as, transform block a size of unit implements the conversion process of difference image that generated by subtraction portion 6 DCT (discrete cosine transform), carry out the KL of bottom-layer design in advance for specific learning series The orthogonal transformation of conversion etc. processes), and implement the following processing: use this coded prediction error to join Several comprised quantization parameters, quantify the conversion coefficient of this difference image, thus will quantify After conversion coefficient export as the compression data of difference image.And, change quantization portion 7 structure Become image compression unit.

Inverse quantization inverse transformation portion 8 implements the following processing: by using from coding-control portion 1 output The quantization parameter that coded prediction error parameter is comprised, to the compression number exported from change quantization portion 7 According to carrying out inverse quantization, the transform block a size of unit comprised with this coded prediction error parameter, real Execute inverse quantization compression data inverse transformation process (such as, inverse DCT (inverse discrete cosine transformation), The inverse transformation of anti-KL conversion etc. processes), thus the compression data after inverse transformation being processed are as office Portion's coding prediction error signal exports.

Adder 9 implements the following processing: by being translated the local exported from inverse quantization inverse transformation portion 8 Code predictive error signal and represent by infra-frame prediction portion 4 or motion compensated prediction portion 5 generate pre- The prediction signal of altimetric image is added, and generates the local decoded picture signal representing local decoded picture.

Infra-frame prediction memorizer 10 is to be represented by the local decoded picture signal generated by adder 9 Local decoded picture as with infra-frame prediction portion 4 next time intra-prediction process in use figure The storage medium of the RAM etc. as preserving.

Loop filter portion 11 implements the following processing: to the local decoded picture generated by adder 9 The coding distortion that signal is comprised compensates, the local decoded picture letter after being compensated by coding distortion Number represent local decoded picture export motion-compensated predicted frames memorizer 12 as with reference to image.

Motion-compensated predicted frames memorizer 12 is the Filtering Processing that will be carried out by loop filter portion 11 After local decoded picture as by motion compensated prediction portion 5 next time motion compensated prediction process The record medium of the RAM etc. preserved with reference to image of middle use.

Variable length code portion 13 implements the following processing: to the compression number exported from change quantization portion 7 According to, from the coding mode of coding-control portion 1 output and coded prediction error parameter, in frame pre- Survey portion 4 output infra-frame prediction parameter or from motion compensated prediction portion 5 output inter prediction ginseng Number carries out variable length code, generates these compression data of multiplexing, coding mode, forecast error volume Code parameter, the bit stream of coded data of infra-frame prediction parameter/inter prediction parameter.And, variable Length coding portion 13 constitutes variable length coding unit.

In fig. 2, variable-length decoder unit 51 implements the following processing: from multiplexing in the bitstream To compression data involved by each encoding block of layering segmentation, coding mode, pre-in coded data Survey error-coding parameters, infra-frame prediction parameter/inter prediction parameter carries out variable-length decoding, by it Compression data and coded prediction error parameter export inverse quantization inverse transformation portion 55, and are compiled Pattern and infra-frame prediction parameter/inter prediction parameter export switching switch 52.And, variable Length decoder unit 51 constitutes variable-length decoding unit.

Switching switch 52 implements the following processing: at the encoding block exported from variable-length decoder unit 51 In the case of involved coding mode is intra-frame encoding mode, will be defeated from variable-length decoder unit 51 The infra-frame prediction parameter gone out exports infra-frame prediction portion 53, when its coding mode is interframe encoding mode In the case of, the inter prediction parameter exported from variable-length decoder unit 51 is exported motion compensation Prediction section 54.

Infra-frame prediction portion 53 implements the following processing: use the picture signal decoded in frame, according to From the infra-frame prediction parameter of switching switch 52 output, by implementing intra-prediction process for encoding block Generate prognostic chart picture.

Wherein, infra-frame prediction internal 53 is after generating above-mentioned prognostic chart picture, from pre-prepd 1 In individual above wave filter, according to generating various parameters known to the moment point of above-mentioned prognostic chart picture State select wave filter, use this wave filter to above-mentioned prognostic chart picture implement Filtering Processing, will Prognostic chart picture after Filtering Processing exports adder 56.

Specifically, as above-mentioned various parameters, according at least 1 in 4 following parameters The state of above parameter determines wave filter uniquely.Wherein, the parameter used predetermine into The parameter identical with dynamic image encoding device described above.That is, it is unified in moving image encoding The parameter used in device and dynamic image decoding device, so that when filling at moving image encoding Put side, in the case of infra-frame prediction portion 4 uses parameter (1) and (4) to be filtered processing, Similarly, infra-frame prediction portion 53 uses parameter (1) and (4) in dynamic image decoding device side It is filtered processing.

Parameter (1)

The block size of above-mentioned prognostic chart picture

Parameter (2)

The quantization parameter of variable-length decoding is carried out by variable-length decoder unit 51

Parameter (3)

Generate prognostic chart as time use frame in the picture signal decoded and Filtering Processing object The distance of pixel

Parameter (4)

The infra-frame prediction parameter of variable-length decoding is carried out by variable-length decoder unit 51

And, switching switch 52 and infra-frame prediction portion 53 constitute intraprediction unit.

Motion compensated prediction portion 54 implements the following processing: use by motion-compensated predicted frames memorizer 59 Reference image more than 1 frame preserved, according to the infra-frame prediction parameters from switching switch 52 output, Prognostic chart picture is generated by encoding block being implemented motion compensated prediction process.And, switching open Pass 52 and motion compensated prediction portion 54 constitute motion compensated prediction unit.

Inverse quantization inverse transformation portion 55 implements the following processing: uses and exports from variable-length decoder unit 51 The quantization parameter that comprised of coded prediction error parameter, to export from variable-length decoder unit 51 Compression data involved by encoding block carry out inverse quantization, by being wrapped with this coded prediction error parameter The transform block contained a size of unit, (such as, the inverse transformation of the compression data implementing inverse quantization processes The inverse transformation of inverse DCT (inverse discrete cosine transformation), anti-KL conversion etc. processes), thus will be anti- Compression data after conversion process (represent the difference image before compression as coding prediction error signal Signal) export.And, inverse quantization inverse transformation portion 55 constitutes difference image signal generating unit.

Adder 56 implements the following processing: by the decoding that will export from inverse quantization inverse transformation portion 55 Predictive error signal and represent by infra-frame prediction portion 53 or motion compensated prediction portion 54 generate pre- The prediction signal of altimetric image is added, and generates the decoded picture signal representing decoded picture.And, add Method portion 56 constitutes decoded picture signal generating unit.

Infra-frame prediction memorizer 57 is to be represented by the decoded picture signal generated by adder 56 Decoded picture is protected as the image used in the intra-prediction process of next time by infra-frame prediction portion 53 The record medium of the RAM etc. deposited.

Loop filter portion 58 implements the following processing: to the decoded picture letter generated by adder 56 Number coding distortion comprised compensates, and the decoded picture signal after being compensated by coding distortion represents Decoded picture as exporting motion-compensated predicted frames memorizer 59 with reference to image.

Motion-compensated predicted frames memorizer 59 is will to carry out at loop filtering with loop filter portion 58 Decoded picture after reason as with motion compensated prediction portion 54 next time motion compensated prediction process in The record medium of the RAM etc. preserved with reference to image used.

In fig. 1, it is assumed that as dynamic image encoding device element coding-control portion 1, Block cutting part 2, switching switch 3, infra-frame prediction portion 4, motion compensated prediction portion 5, subtraction portion 6, Change quantization portion 7, inverse quantization inverse transformation portion 8, adder 9, loop filter portion 11 and can With special hardware, (quasiconductor such as, being provided with CPU is integrated respectively for elongated degree encoding section 13 Circuit or single-chip microcomputer etc.) constitute, but at dynamic image encoding device with calculating the feelings that mechanism becomes Under condition, it is also possible to pre-in record is had coding-control portion 1, block cutting part 2, switching switch 3, frame Survey portion 4, motion compensated prediction portion 5, subtraction portion 6, change quantization portion 7, inverse quantization inverse transformation portion 8, the journey of the process content in adder 9, loop filter portion 11 and variable length code portion 13 Sequence is stored in the memorizer of this computer, and the CPU of this computer performs to be stored in the journey of this memorizer Sequence.

In FIG. 2, it is assumed that the variable-length decoding of the element as dynamic image decoding device Portion 51, switching switch 52, infra-frame prediction portion 53, motion compensated prediction portion 54, inverse quantization contravariant Change portion 55, adder 56 and loop filter portion 58 respectively by special hardware (such as, peace Equipped with the semiconductor integrated circuit of CPU or single-chip microcomputer etc.) constitute, but at moving image decoding dress Put in the case of becoming with calculating mechanism, it is also possible to record has variable-length decoder unit 51, switching open Pass 52, infra-frame prediction portion 53, motion compensated prediction portion 54, inverse quantization inverse transformation portion 55, addition The program of the process content in portion 56 and loop filter portion 58 is stored in the memorizer of this computer, The CPU of this computer performs to be stored in the program of this memorizer.

Following description action.

Initially, the process content of the dynamic image encoding device of explanatory diagram 1.

First, coding-control portion 1 determine implement intra-prediction process (intra-prediction process) or The maximum becoming the encoding block processing unit when motion compensated prediction processes (inter prediction process) Size, and determine to be layered the upper limit number of plies (step of Fig. 3 when splitting maximum sized encoding block ST1)。

As the maximum sized determining method of encoding block, such as, it is considered for whole pictures, certainly The method making size corresponding with the resolution of input picture.

Additionally, it is contemplated that using the difference of the complexity of the motion of the locality of input picture as parameter Carry out quantification, full-size is determined as little value by the picture of motion intense, few in motion Picture in full-size is determined as the method etc. of big value.

Setting for the number of plies of the upper limit such as considers following method, i.e., is set as input picture Motion violent, all the more deep layer number, so that trickleer motion can be detected, if input figure The motion of picture is few, then suppress the number of plies.

It addition, coding-control portion 1 is from the coding mode of utilizable more than 1 (in M kind frame Coding mode, N kind interframe encoding mode) in select to be suitable for each encoding block of layering segmentation Coding mode (step ST2).

The system of selection of coding mode based on coding-control portion 1 because be known technology, so Omit detailed description, for instance that use utilizable arbitrary coding mode, to encoding block Implement coded treatment, verify code efficiency, in utilizable multiple coding modes, select to compile The method etc. of the coding mode that code efficiency is best.

It addition, each to each encoding block of coding-control portion 1, determine to make during compressed differential image Quantization parameter and transform block size, and the frame determining to implement to use when prediction processes in pre- Survey parameter or inter prediction parameter.

Coding-control portion 1 will comprise the coded prediction error parameter of quantization parameter and transform block size Change quantization portion 7, inverse quantization inverse transformation portion 8 and variable length code portion 13 are arrived in output.It addition, As required coded prediction error parameter is exported infra-frame prediction portion 4.

If block cutting part 2 input represents the video signal of input picture, then by this video signal table The input picture shown is divided into the maximum sized encoding block determined by coding-control portion 1, and directly Till reaching the upper limit number of plies with coding-control portion 1 decision, this encoding block is split in layering.

Here, Fig. 5 is the sample representing and maximum sized encoding block being divided into multiple encoding block The explanatory diagram of son.

In the example of fig. 5, maximum sized encoding block is the encoding block B of the 0th layer⁰, according to bright Degree composition has (L⁰, M⁰) size.

It addition, in the example of fig. 5, by maximum sized encoding block B⁰As initial point, with 4 forks Tree construction is divided into the prescribed depth being otherwise determined that, thus obtains encoding block Bⁿ。

In degree of depth n, encoding block BⁿIt is size (Lⁿ、Mⁿ) image-region.

Wherein, LⁿAnd MⁿCan be the same or different, but L is shown in the example of fig. 5ⁿ=Mⁿ Situation.

After, encoding block BⁿSize be defined as encoding block BⁿLuminance components in (Lⁿ, Mⁿ)。

Block cutting part 2 is because carrying out 4 fork tree segmentations, so (Lⁿ⁺¹, Mⁿ⁺¹)=(Lⁿ/ 2, Mⁿ/2) All the time set up.

Wherein, as rgb signal etc., there is at whole color components the coloured silk of same sample number In color video signal (4:4:4 form), the size of whole color components becomes (Lⁿ, Mⁿ), But in the case of processing 4:2:0 form, the size of the encoding block of corresponding aberration composition is (Lⁿ/ 2, Mⁿ/2)。

After, by the encoding block B in n-th layerⁿIn the coding mode that can select be labeled as m (Bⁿ)。

In the case of being the colour-video signal with multiple color component, coding mode m (Bⁿ) Each color composition can be constituted in the way of using independent pattern respectively, as long as but do not have later There is special instruction, it is simply that be set to refer to the luminance components of the encoding block to YUV signal 4:2:0 form The situation of coding mode illustrate.

At coding mode m (BⁿOne or more intra-frame encoding modes are had (to be generically and collectively referred to as in) " INTRA "), one or more interframe encoding modes (being generically and collectively referred to as " INTER "), coding-control Portion 1 is as it has been described above, select from whole coding mode utilizable this picture or its subset Select encoding block BⁿThe coding mode that code efficiency is best.

Encoding block Bⁿ(divide as it is shown in figure 5, be further divided into one or more predictions process unit District).

After, encoding block B will be belonged toⁿLabelling Regions be P_i ⁿ(area code of i: n-th layer).

Encoding block B will be belonged toⁿSubregion P_i ⁿSegmentation how to carry out being included in coding as information Mode m (BⁿIn).

Subregion P_i ⁿAll according to coding mode m (Bⁿ) be predicted processing, but for each subregion P_i ⁿ, independent Prediction Parameters can be selected.

Coding-control portion 1 for maximum sized encoding block, such as, generates block as shown in Figure 6 Cutting state, determines encoding block Bⁿ。

The dash area of Fig. 6 (a) represents the distribution of the subregion after segmentation, it addition, Fig. 6 (b) uses 4 fork tree graphs represent coding mode m (Bⁿ) distribute to be layered the situation of the subregion after splitting.

In Fig. 6 (b), represent with the node of square encirclement and coding mode m (B is assignedⁿ) Node (encoding block Bⁿ)。

Select each encoding block B in coding-control portion 1ⁿSubregion P_i ⁿOptimal coding mode m (Bⁿ) time, if this coding mode m (Bⁿ) be intra-frame encoding mode (step ST3), then The encoding block B that switching switch 3 will have been split with block cutting part 2ⁿSubregion P_i ⁿOutput is to pre-in frame Survey portion 4.

On the other hand, if this coding mode m (Bⁿ) it is interframe encoding mode (step ST3), The encoding block B then will split with block cutting part 2ⁿSubregion P_i ⁿOutput is to motion compensated prediction portion 5。

Encoding block B is being received from switching switch 3ⁿSubregion P_i ⁿTime, intra-coding prediction portion 4 Use the encoded picture signal in frame, according to the infra-frame prediction ginseng exported from coding-control portion 1 Number, by this encoding block BⁿSubregion P_i ⁿImplement intra-prediction process, prognostic chart in delta frame As P_i ⁿ(step ST4).

But, infra-frame prediction portion 4 is generating above-mentioned intra-prediction image P_i ⁿAfter, from pre-prepd In the wave filter of more than 1, generating and above-mentioned infra-frame prediction according in dynamic image decoding device Image P_i ⁿKnown to the moment point of identical prognostic chart picture, the state of various parameters selects wave filter, Use this wave filter to this intra-prediction image P_i ⁿImplement Filtering Processing.

Infra-frame prediction portion 4 is when to intra-prediction image P_i ⁿWhen implementing Filtering Processing, after Filtering Processing Intra-prediction image P_i ⁿOutput is to subtraction portion 6 and adder 9, even and if in the motion of Fig. 2 Also because identical intra-prediction image P can be generated in image decoder_i ⁿ, so by this frame Prediction Parameters exports variable length code portion 13.

The outline of the process content in infra-frame prediction portion 4 is as it has been described above, process explanation after content in detail.

Motion compensated prediction portion 5 is receiving encoding block B from switching switch 3ⁿSubregion P_i ⁿTime, Use the reference image of more than 1 frame with motion-compensated predicted frames memorizer 12 preservation, according to from volume The inter prediction parameter of code control portion 1 output, by this encoding block BⁿSubregion P_i ⁿImplement fortune Dynamic compensation prediction processes, and generates interframe prognostic chart as P_i ⁿ(step ST5).

And, process the technology generating prognostic chart picture because of being public affairs by implementing motion compensated prediction The technology known is so omitting detailed description.

Prognostic chart picture (infra-frame prediction figure is generated in infra-frame prediction portion 4 or motion compensated prediction portion 5 As P_i ⁿ, inter prediction image P_i ⁿ) time, subtraction portion 6 is by from the volume split with block cutting part 2 Code block BⁿSubregion P_i ⁿIn deduct and generate with infra-frame prediction portion 4 or motion compensated prediction portion 5 Prognostic chart picture (intra-prediction image P_i ⁿ, inter prediction image P_i ⁿ) generate difference image, by table Show the predictive error signal e of this difference image_i ⁿOutput is to change quantization portion 7 (step ST6).

Change quantization portion 7 represents the predictive error signal e of difference image from subtraction portion 6 receiving_i ⁿ Time, the transform block comprised with the coded prediction error parameter exported from coding-control portion 1 is a size of Unit, implements conversion process (such as, the DCT (discrete cosine transform), pre-of this difference image The orthogonal transformation of the KL conversion etc. that specific learning series first carries out bottom-layer design processes), and And use the quantization parameter being contained in this coded prediction error parameter, the transformation series to this difference image Number quantifies, thus using the conversion coefficient after this quantization as the compression data of difference image, defeated Go out to inverse quantization inverse transformation portion 8 and variable length code portion 13 (step ST7).

Inverse quantization inverse transformation portion 8 is receiving the compression data of difference image from change quantization portion 7 Time, use the quantization parameter that the coded prediction error parameter from coding-control portion 1 output is comprised, The compression data of this difference image are carried out inverse quantization, to be contained in this coded prediction error parameter Transform block a size of unit, the inverse transformation of the compression data implementing inverse quantization processes (such as, inverse DCT The inverse transformation of (inverse discrete cosine transformation), anti-KL conversion etc. processes), thus by inverse transformation Compression data after reason are as local coding prediction error signal e_i ⁿCap (hat) is (because of electronic application Relation, " ^ " that be attached on alphabetic character is designated as cap), output to adder 9 (step ST8).

Adder 9 is receiving local coding prediction error signal e from inverse quantization inverse transformation portion 8_i ⁿCap Time, by by this local coding prediction error signal e_i ⁿCap and expression infra-frame prediction portion 4 or fortune Prognostic chart picture (intra-prediction image P that dynamic compensation prediction portion 5 generates_i ⁿ, inter prediction image P_i ⁿ) Prediction signal be added, generate local decoded picture, this local decoded picture be local decode subregion Image P_i ⁿCap or as its local gathered decoding encoding block image (step ST9).

Adder 9, when generating local decoded picture, would indicate that the local of this local decoded picture is translated Code picture signal is stored in infra-frame prediction memorizer 10, and by defeated for this local decoded picture signal Go out to loop filter portion 11.

The process of step ST3～ST9 is at the whole encoding block B to layering segmentationⁿProcess terminate Repeat to implement, when to whole encoding block B beforeⁿProcess at the end of transfer to the place of step ST12 Reason (step ST10, ST11).

The variable length code portion 13 compression data to exporting from change quantization portion 7, from coding-control The coding mode (comprising the information of the cutting state of presentation code block) of portion 1 output and prediction are by mistake Differ from coding parameter, from the infra-frame prediction parameter of infra-frame prediction portion 4 output or from motion compensated prediction The inter prediction parameter of portion 5 output carries out entropy code.

Variable length code portion 13 multiplexing as the coding result of entropy code compression data, coding mould Formula, coded prediction error parameter, the coded data of infra-frame prediction parameter/inter prediction parameter, generate Bit stream (step ST12).

Loop filter portion 11 is when receiving local decoded picture signal from adder 9, and compensation comprises In the coding distortion of this local decoded picture signal, the local decoded picture after coding distortion is compensated The local decoded picture that signal represents is as being stored in motion-compensated predicted frames memorizer with reference to image 12 (steps ST13).

The Filtering Processing carried out by loop filter portion 11 both can be with the local exported from adder 9 The maximum encoding block of decoded picture signal or respective encoding block are that unit is carried out, it is also possible to converge The unit collecting multiple maximum encoding block is carried out, or the local decoding figure in output 1 picture amount 1 picture amount of collecting after image signal is carried out.

Then the process content in infra-frame prediction portion 4 is described in detail.

Represent the prediction side that intra prediction mode and this intra prediction mode represent in the example of fig. 7 To vector, it is designed to the increase of number along with the intra prediction mode that can select, it was predicted that direction Vector relative angle each other reduces.

Infra-frame prediction portion 4 is according to for subregion P_i ⁿInfra-frame prediction parameter, in intra-prediction image P_i ⁿGeneration in the Selection parameter of wave filter that uses, to subregion P_i ⁿImplement intra-prediction process.

Hereinafter, illustrate according to for subregion P_i ⁿThe infra-frame prediction parameter (infra-frame prediction of luminance signal Pattern), process in the frame of the infra-frame prediction signal generating luminance signal.

In this, it is assumed that subregion P_i ⁿSize be l_i ⁿ×m_i ⁿPixel.

Fig. 8 is to represent at l_i ⁿ=m_i ⁿWhen=4, generating subregion P_i ⁿIn pixel predictors time use The explanatory diagram of an example of pixel.

In fig. 8, will be with subregion P_i ⁿPixel (2 × the l of adjacent encoded upper subregion_i ⁿ+ 1) individual With the pixel (2 × m parting district on the left side_i ⁿ) the individual pixel being set to for prediction, but the picture used in prediction Element is the amount doesn't matter compared with the pixel shown in Fig. 8.

It addition, in fig. 8, although in prediction, use 1 adjacent row or the pixel of 1 row amount, But 2 row or 2 row can also be used in prediction, or its above pixel.

When for subregion P_i ⁿThe index value of intra prediction mode be the situation of 2 (meansigma methods predictions) Under, using the meansigma methods of the neighbor of upper subregion and the neighbor parting district on the left side as subregion P_i ⁿIn All the predictive value of pixel generates medium range forecast image.

In the case of intra prediction mode index value is beyond 2 (meansigma methods predictions), according to subregion The prediction direction vector v that value represents_p=(dx, dy) generates subregion P_i ⁿThe predictive value of interior pixel.

Where it is assumed that generate the subregion P of the pixel (prediction subject pixels) of predictive value_i ⁿInterior is relative Coordinate (with the top left pixel of subregion as initial point) be (x, y).

The position of reference pixels used in prediction becomes at the A of following expression and neighbor Intersection point.

A = (\begin{matrix} x \\ y \end{matrix}) + {k&upsi;}_{p}

Wherein, k is negative scalar value.

In the case of reference pixels is in integer pixel positions, it is right to be set to this integer pixel to predict Predictive value as pixel.

On the other hand, in the case of reference pixels is not at integer pixel positions, will according to ginseng The interpolating pixel that the integer pixel that the element that takes pictures is adjacent generates is set to predictive value.

In the example of fig. 8, because reference pixels is not at integer pixel positions, thus according to 2 pixels that reference pixels is adjacent carry out interpolation, calculate predictive value.But, it was predicted that value does not limit In 2 adjacent pixels, it is also possible to generate interpolating pixel according to the pixel of adjacent more than 2 And as predictive value.

Then, by the subregion P generated according to above-mentioned steps_i ⁿThe centre that interior predictive value is constituted Prognostic chart Filtering Processing described in after carrying out, obtains final intra-prediction image P_i ⁿ, should Intra-prediction image P_i ⁿExport in subtraction portion 6 and adder 9.

It addition, in intra-prediction image P_i ⁿGeneration in use infra-frame prediction parameter in order to multiplexing in Bit stream and export variable length code portion 13.

Hereinafter, concrete Filtering Processing is described.

From at least more than one wave filter pre-prepd, with the method choice institute of explanation later The wave filter used, is filtered according to below formula (1) each pixel of middle prognostic chart picture Process.

\hat{s} (p_{0}) = a_{0} s (p_{0}) + a_{1} s (p_{1}) + ... + a_{N - 1} s (p_{N - 1}) + a_{N} - - - (1)

In formula (1), a_n(n=0,1 ..., N) it is to comprise the coefficient that reference pixels relates to (a₀、a₁、……、a_N-1) and deviation ratio a_NFilter coefficient.

P_n(n=0,1 ..., N-1) represent comprise Filtering Processing object p₀The ginseng of wave filter Take pictures element.N is arbitrary reference pixels number.

s(p_n) represent the brightness value of each reference pixels, s cap (p₀) represent at Filtering Processing object Pixel p₀In Filtering Processing after brightness value.

But, filter coefficient can also be set to do not have deviation ratio a_NCoefficient constitute.It addition, For being in subregion P_i ⁿBrightness value s (the p of interior each reference pixels_n), both can be set to middle pre- The brightness value of each pixel of altimetric image, it is also possible to the location of pixels being set to only terminate in Filtering Processing enters Brightness value after row Filtering Processing.

For being in subregion P_i ⁿOutside the brightness value s (p of each reference pixels_n), if compiling Code region, then be the brightness value (decoded brightness value) after coding, if the most do not carried out The region of coding, then from being in subregion P defined above_i ⁿBrightness value s (the p of interior each reference pixels_n) And order according to regulation selects alternative signal value in the brightness value after the coding in encoded region (such as, as in the pixel of candidate, selecting to be positioned at the signal value etc. of nearest position).

When carrying out above-mentioned Filtering Processing, subregion P_i ⁿSize (l_i ⁿ×m_i ⁿ) the biggest, in input Image the most easily exists nonlinear edge etc., because be susceptible to medium range forecast image is pre- Survey the deviation in direction, it is advantageous to middle prognostic chart picture is smoothed.

And then, it was predicted that the quantized value of error is the biggest, and the quantizing distortion produced in decoded picture is the biggest, With subregion P according to Yin_i ⁿAdjacent encoded pixel and the prediction of medium range forecast image that generates Precision reduces, so the standby performance subregion P roughly of preference level_i ⁿAs pre-through smoothing processing Altimetric image.

And then, even identical subregion P_i ⁿIn pixel because distance with at medium range forecast image Generation in use subregion P_i ⁿThe pixel that adjacent encoded pixel is the most remote, the easiest to be pre-in centre The skew at edge etc. is there is, so prognostic chart picture to be carried out smooth place between altimetric image and input picture Reason, suppresses the increase at the violent forecast error occurred when offseting relatively good.

It addition, infra-frame prediction when generating medium range forecast image is configured to, according in prediction block Predictive value is the meansigma methods prediction of the most identical value and uses prediction direction vector v_pPrediction this 2 The method planting one of different method is predicted, even and if then using prediction direction vector v_p's In the case of prediction, also according to will be located in the value of reference pixels of integer pixel positions directly as in advance The reference pixels more than pixel of measured value and at least 2 pixels carries out interpolation, and generation is not positioned at whole Number location of pixels pixel, using this pixel value as predictive value pixel prediction block in configuration because of Prediction direction vector v_pDirection and different.Thus, it was predicted that the character of image is because of intra prediction mode And different, optimal Filtering Processing is the most different, so changing according to the index value of intra prediction mode Become the intensity of wave filter, the reference pixels number of wave filter, reference pixels configuration etc. relatively good.

Thus, in wave filter selection processes, it is configured to 4 parameters (1) below considering～(4) Select wave filter.

(1) subregion P_i ⁿSize (l_i ⁿ×m_i ⁿ)

(2) quantization parameter of coded prediction error parameter it is contained in

(3) when the generation of medium range forecast image use encoded pixel (shown in Fig. 8 " The pixel used in prediction ") group and the distance of Filtering Processing subject pixels

(4) index value of intra prediction mode during generation medium range forecast image

Specifically it is configured to: subregion P_i ⁿSize (l_i ⁿ×m_i ⁿ) the biggest, quantization parameter determine Fixed quantized value is the biggest, Filtering Processing subject pixels and be in subregion P_i ⁿThe left side and top The distance of the pixel group of coding is the most remote, more uses wave filter or reference pixels that smoothing processing intensity is strong The wave filter that number is many.As Filtering Processing subject pixels and be in subregion P_i ⁿThe left side and top One example of the distance of encoded pixel group, lists Figure 11.It addition, be configured to also utilize The index value of intra prediction mode comes the intensity of switching filter, the reference pixels number of wave filter, ginseng Take pictures element configuration etc..

That is, for each combination of above-mentioned parameter combination, by from pre-prepd flora of filters In carry out the correspondence of suitable wave filter, it is achieved the adaptability choosing of wave filter corresponding with above-mentioned parameter Select.It addition, such as, when combination parameter (3) and (4), it is also possible to according to parameter (4) " intra prediction mode " changes the definition of " with the distance of encoded pixel group " of parameter (3) aptly. That is, it is not as shown in Figure 11 by the definition immobilization with the distance of encoded pixel group, yet Can be set to away from the distance depending on prediction direction as the distance of " reference pixels " shown in Fig. 8. By so processing, it is possible to realize being additionally contemplates that parameter (3) and (4) such multiple parameters are closed The adaptive Filtering Processing of system.

It addition, in the combination of these parameters, can prepare accordingly not with " there is no Filtering Processing " Implement the combination of Filtering Processing.And then, as the definition of the intensity of wave filter, it is also possible to will be the most weak Wave filter be defined as " there is no Filtering Processing ".

It addition, 4 parameters (1)～(4) are even if because in dynamic image decoding device side also It is known parameter, so the most not producing for should compile required for carrying out above-mentioned Filtering Processing The additional information of code.

And, in the above description, carry out adaptability by preparing the wave filter of necessary amount in advance Select, thus be filtered the switching of device but it also may by select ginseng according to above-mentioned wave filter Wave filter is defined as the function of above-mentioned wave filter Selection parameter by the mode of the value calculating wave filter of number, To realize the switching of wave filter.

And then, in the above description, illustrate and be configured to consider 4 parameters (1)～(4) Select the situation of wave filter but it also may be configured in 4 parameters (1)～(4), it is considered to The parameter of more than at least 1 selects wave filter.

Hereinafter, the situation of parameter (1) and (4) will be used as an example, represent by ginseng Array close each combination make suitable filtering be mapped from pre-prepd flora of filters thus The composition example of the Filtering Processing that the adaptability being filtered selects.

The wave filter used in above-mentioned Filtering Processing example is defined as described below.

The wave filter (reference pixels number N=3) of filter index 1

a₀=3/4, a₁=1/8, a₂=1/8

The wave filter (reference pixels number N=3) of filter index 2

a₀=1/2, a₁=1/4, a₂=1/4

The wave filter (reference pixels number N=3) of filter index 3

a₀=1/4, a₁=3/8, a₂=3/8

The wave filter (reference pixels number N=5) of filter index 4

a₀=1/4, a₁=3/16, a₂=3/16, a₃=3/16, a₄=3/16

But, it is assumed that Filtering Processing is set to do not have deviation ratio a_N(a_N=0) time based on formula (1) Filtering Processing, it is assumed that the kind of the wave filter this time used is 3 kinds, the reference pixels of wave filter Configuration is as shown in figure 12.

Then, Figure 13 is to illustrate that at subregion P_i ⁿEach size each intra prediction mode in The explanatory diagram of one example of the table of the wave filter used.Where it is assumed that the P that can obtain_i ⁿChi Very little is 4 × 4 pixels, 8 × 8 pixels, 16 × 16 pixels, 32 × 32 pixels, 64 × 64 pixels, in frame The index value of predictive mode and the correspondence of intra prediction direction are as shown in Figure 7.

It is not filtered processing it addition, filter index 0 represents.Typically when use directional prediction, In the case of meansigma methods prediction, because there is following trend, so table is such as shown in fig. 13 that, The characteristic of the image in consideration infra-frame prediction, by each combination to parameter (1) and (4), Which wave filter of use is mapped by table such that it is able to it is suitable to realize by referring to table Wave filter switches.

The general line at the edge in most horizontal vertical directions existed in the cultures such as building Property situation clearly is more, so most cases can carry out high-precision by horizontal vertical direction prediction The prediction of degree.Thus it is not smoothed relatively good when horizontal vertical direction prediction.

Generally because picture signal is at the high signal of direction in space seriality, thus use with and Subregion P_i ⁿDuring the impaired meansigma methods prediction of the seriality of adjacent encoded pixel, to subregion P_i ⁿ's The pixel of the block boundary portion periphery of the left side and top is smoothed improving seriality and compares Good.

Having the region of the directivity of incline direction because area is the biggest, edge etc. occurs distortion to have The situation having non-linear shape is the most, so partitions sizes is the biggest when using incline direction prediction, More apply the wave filter that smoothed intensity is strong, reference pixels number is many relatively good.

If general partitions sizes is excessive, then the spatial variations of the signal value in subregion becomes various, The most rough prediction can only be carried out, it is difficult to high-precision in directional prediction, meansigma methods prediction The region of prediction increases.In such region because be smoothed be only make image blurring And the improvement effect of predictive efficiency cannot be expected, so in such partitions sizes, because not The side implementing Filtering Processing does not increase unnecessary amount of calculation, so being (the most such as, In fig. 13, the partitions sizes more than 32 × 32 pixels is set to be not filtered process).

And then, the reference pixels when in Filtering Processing is subregion P_i ⁿIn pixel time by medium range forecast The brightness value of image is set in the case of the brightness value of reference pixels uses, sometimes to simplify filter Ripple processes.Such as, in the case of intra prediction mode is meansigma methods prediction, for subregion P_i ⁿ's Filtering Processing can be simplified to following such Filtering Processing for each region shown in Figure 14.

Region A (subregion P_i ⁿThe pixel of upper left)

The wave filter (not changing) of filter index 1

a₀=3/4, a₁=1/8, a₂=1/8 (reference pixels number N=3)

The wave filter (not changing) of filter index 2

a₀=1/2, a₁=1/4, a₂=1/4 (reference pixels number N=3)

The wave filter (not changing) of filter index 3

a₀=1/4, a₁=3/8, a₂=3/8 (reference pixels number N=3)

The wave filter of filter index 4

a₀=5/8, a₁=3/16, a₂=3/16 (reference pixels number N=3)

Region B (the subregion P beyond the A of region_i ⁿThe pixel of upper end)

The wave filter of filter index 1

a₀=7/8, a₂=1/8 (reference pixels number N=2)

The wave filter of filter index 2

a₀=3/4, a₂=1/4 (reference pixels number N=2)

The wave filter (not changing) of filter index 3

a₀=5/8, a₂=3/8 (reference pixels number N=2)

The wave filter of filter index 4

a₀=13/16, a₂=3/16 (reference pixels number N=2)

Region C (the subregion P beyond the A of region_i ⁿThe pixel of left end)

The wave filter of filter index 1

a₀=7/8, a₁=1/8 (reference pixels number N=2)

The wave filter of filter index 2

a₀=3/4, a₁=1/4 (reference pixels number N=2)

The wave filter of filter index 3

a₀=5/8, a₁=3/8 (reference pixels number N=2)

The wave filter of filter index 4

a₀=13/16, a₁=3/16 (reference pixels number N=2)

Region D (the subregion P beyond region A, B, C_i ⁿPixel)

The wave filter of whole filter index:

There is no Filtering Processing

Even if simplifying Filtering Processing as described above, Filtering Processing result is also identical with before simplification.

By so removing the redundancy section of actual treatment, it is possible to seek the high speed of Filtering Processing.

Employ the table of Figure 13 in the above example but it also may use other table.Such as, also Can consider when the situation paying attention to suppression treating capacity compared with the size improving effect of coding efficiency Under, replace the table of Figure 13 to use the table of Figure 19.In the case of this table, because only to subregion P_i ⁿSize be 4 × 4 pixels, 8 × 8 pixels, 16 × 16 pixels meansigma methods prediction be filtered place Reason, so the predictive mode being filtered processing compared with the situation of the table using Figure 13 is few, it is possible to Suppression is with the increase of the amount of calculation of Filtering Processing.Now, by also utilizing above-mentioned infra-frame prediction Pattern is the simplification of Filtering Processing during meansigma methods prediction, it is possible to realization processes the low-down filter of load Ripple processes.

And then, in the case of also paying attention to the easiness implemented, the same false with above-mentioned Filtering Processing If being only filtered processing, not according to subregion P in the case of meansigma methods is predicted_i ⁿSize switch The wave filter used further, can use same wave filter (such as, filter index 2 all the time Wave filter).In this case, although the improvement effect of coding efficiency based on wave filter reduces With do not carry out and subregion P_i ⁿSize process corresponding amount accordingly, but can suppress when implementing Circuit scale (being the line number of code in the case of software).This Filtering Processing is only to consider 4 The wave filter of the parameter (4) in parameter (1)～(4).

As the embodiment of Filtering Processing, even if not to select the filtering of correspondence by referring to table The form of the wave filter of device index is implemented, it is also possible to be directly to implement subregion P_i ⁿEach size Perform Filtering Processing or directly enforcement to subregion P_i ⁿEach size each location of pixels perform The form of this direct enforcement wave filter of Filtering Processing.Such that make the form rather than referring to table, As long as it seems of equal value for being filtered the prognostic chart obtained by the result processed, then embodiment is All it doesn't matter for which kind of form.

It addition, in example described above, illustrate only with a table for switching filter Method but it also may be configured to prepare multiple above-mentioned table, a certain by with in Figure 15～Figure 18 Wave filter is selected table index 100 to encode as heading message by individual form, such that it is able to use The unit of regulation carrys out switching filter selection table.

Such as, as shown in figure 15, table index 100 is selected by sequence-level title being added upper filter, Compared with the situation only using single table, it is also possible to carry out at filtering corresponding with the characteristic of sequence Reason.

And, with the block of 8 × 8 pixels in MPEG-4, AVC/H.264 described above Smoothing processing during infra-frame prediction implemented reference image is the same, even being configured to pre-in frame Survey portion 4 generates subregion P_i ⁿMedium range forecast image time reference pixels as to subregion P_i ⁿPhase In the case of adjacent encoded pixel has carried out the pixel of smoothing processing, it is also possible to carry out and above-mentioned example The Filtering Processing to middle prognostic chart picture that son is the same.

On the other hand, because based on the smoothing processing to reference pixels when generating medium range forecast image With the part that the effect of the Filtering Processing to middle prognostic chart picture has repetition, even if so sometimes simultaneously Use two process, also can only obtain a little property compared with the situation carrying out the wherein process of a side Effect can be improved.Therefore, it is possible to be configured to: in the case of paying attention to suppression amount of calculation, for right The subregion P that reference pixels when generating medium range forecast image is smoothed_i ⁿ, do not carry out for The Filtering Processing of medium range forecast image.For example, it is contemplated that such situation: to middle prognostic chart as Filtering Processing is only filtered processing, for generating as the table of Figure 19 in meansigma methods is predicted The smoothing processing reference of the reference pixels of medium range forecast image is only to directivity specific as Figure 20 The table that prediction is smoothed is carried out.Wherein, in fig. 20, ' 1 ' expression is smoothed, ' 0 ' expression is not smoothed.

In intra-prediction image P_i ⁿGeneration in use infra-frame prediction parameter in order to multiplexing is in bit In stream, output is to variable length code portion 13.

To subregion P_i ⁿInterior colour difference signal, also according to the order same with luminance signal, is implemented based on frame The intra-prediction process of interior Prediction Parameters (intra prediction mode), by the life in intra-prediction image The infra-frame prediction parameter used in one-tenth exports variable length code portion 13.

But, the infra-frame prediction about colour difference signal had both been configured to enter in the same manner as luminance signal Row is in the Filtering Processing of described above, it is also possible to do not carry out the Filtering Processing in described above.

Then, the process content of the dynamic image decoding device of explanatory diagram 2.

Variable-length decoder unit 51 when the bit stream that input exports from the picture coding device of Fig. 1, This bit stream is implemented variable-length decoding process, with the sequence of the picture with more than 1 frame as list Position or the information (step ST21 of Fig. 4) being unit decoded frame size with picture.

Variable-length decoder unit 51 determines in reality according to the step the same with the coding-control portion 1 of Fig. 1 Execute intra-prediction process (intra-prediction process) or motion compensated prediction processes (at inter prediction Reason) time become the full-size of encoding block processing unit, and determine the maximum chi of layering segmentation The upper limit number of plies (step ST22) during very little encoding block.

Such as, in picture coding device, when full-size the dividing according to input picture of encoding block In the case of resolution is determined, determine the maximum of encoding block according to the frame size information decoded before Size.

And, the full-size of presentation code block and the upper limit number of plies when multiplexing in the bitstream In the case of information, with reference to the information decoded according to this bit stream.

It is being multiplexed to the maximum sized encoding block B of bit stream⁰Coding mode m (B⁰In), Comprise the maximum sized encoding block B of expression⁰The information of cutting state, so variable-length decoder unit 51 decodings are multiplexed to the maximum sized encoding block B of bit stream⁰Coding mode m (B⁰), really Each encoding block B of fixed layering segmentationⁿ(step ST23).

Variable-length decoder unit 51 is when determining each encoding block BⁿTime, decode this encoding block BⁿVolume Pattern m (Bⁿ), according to belonging to this coding mode m (Bⁿ) subregion P_i ⁿInformation, determines genus In encoding block BⁿSubregion P_i ⁿ。

Variable-length decoder unit 51 belongs to encoding block B when determiningⁿSubregion P_i ⁿTime, for each point District P_i ⁿ, decoding compression data, coding mode, coded prediction error parameter, infra-frame prediction parameter/ Inter prediction parameter (step ST24).

That is, when distributing to encoding block BⁿCoding mode m (Bⁿ) it is the situation of intra-frame encoding mode Under, to each subregion P belonging to encoding block_i ⁿPrediction Parameters in decoded frame.

When distributing to encoding block BⁿCoding mode m (Bⁿ) be interframe encoding mode in the case of, To each subregion P belonging to encoding block_i ⁿDecoding interframe Prediction Parameters.

Become prediction and process the change that the subregion of unit is comprised according to coded prediction error parameter further Change block dimension information, be divided into one or more subregions becoming conversion process unit, to becoming Each subregion decoding compression data (conversion coefficient after change quantization) of conversion process unit.

Encoding block B is being belonged to from variable-length decoder unit 51ⁿSubregion P_i ⁿCoding mode m (Bⁿ) it is in the case of intra-frame encoding mode (step ST25), switching switch 52 will be from variable The infra-frame prediction parameter of length decoder unit 51 output exports infra-frame prediction portion 53.

On the other hand, as subregion P_i ⁿCoding mode m (Bⁿ) be interframe encoding mode in the case of (step ST25), exports fortune by the inter prediction parameter exported from variable-length decoder unit 51 Dynamic compensation prediction portion 54.

Infra-frame prediction portion 53 is when receiving infra-frame prediction parameter from switching switch 52, and Fig. 1 As infra-frame prediction portion 4, use the picture signal decoded in frame, join according to this infra-frame prediction Number, to encoding block BⁿSubregion P_i ⁿImplement intra-prediction process, generate intra-prediction image P_i ⁿ(step Rapid ST26).

But, infra-frame prediction portion 53 is generating above-mentioned intra-prediction image P_i ⁿAfter, with Fig. 1's The method that infra-frame prediction portion 4 is the same, from the wave filter of pre-prepd more than 1, according to Generate above-mentioned intra-prediction image P_i ⁿMoment point known to the state of various parameters select filtering Device, uses this wave filter to this intra-prediction image P_i ⁿImplement Filtering Processing, after Filtering Processing Intra-prediction image P_i ⁿAs final intra-prediction image.

That is, by the parameter identical with the parameter of use in the wave filter selection in infra-frame prediction portion 4, Carry out with the method choice wave filter the same with the wave filter system of selection in infra-frame prediction portion 4 Filtering Processing.

Such as, infra-frame prediction portion 4 is configured to when situation and the wave filter not being filtered processing Index 0 is mapped, by right with filter index 1 to 4 respectively for pre-prepd 4 wave filter Should get up, in the case of being filtered processing with reference to the table of Figure 13, infra-frame prediction portion 53 also defines The wave filter the same with infra-frame prediction portion 4 and filter index, by referring to Figure 13 table according to point District P_i ⁿSize and index as the intra prediction mode of infra-frame prediction parameter be filtered device Select, be filtered processing.

It addition, example described above is such, it is also possible to be configured to: preparing the definition group with parameter It is combined into the table of the wave filter that unit uses, realizes the situation of the switching of wave filter by referring to this table Under, select table index 100 as mark in wave filter according to the form of some in Figure 15～Figure 18 Topic information decodes, and selects from the table group identical with pre-prepd dynamic image encoding device Decoded wave filter selects the table that table index 100 represents, is filtered the choosing of device with reference to this table Select.

Motion compensated prediction portion 54, when receiving inter prediction parameter from switching switch 52, uses Reference image more than 1 frame preserved by motion-compensated predicted frames memorizer 59, according to this interframe Prediction Parameters, by encoding block B_nSubregion P_i ⁿImplement motion compensated prediction to process, generate Inter prediction image P_i ⁿ(step ST27).

Inverse quantization inverse transformation portion 55 uses the coded prediction error from variable-length decoder unit 51 output The quantization parameter that parameter is comprised, the pressure that the encoding block exported from variable-length decoder unit 51 is related to Contracting data carry out inverse quantization, the transform block a size of unit comprised with this coded prediction error parameter, (such as, inverse DCT (anti-discrete cosine is processed by the inverse transformation implementing the compression data of inverse quantization Conversion), the inverse transformation of anti-KL conversion etc. processes), the compression data after inverse transformation being processed are made For coding prediction error signal (representing the signal of the difference image before compression) output to adder 56 (step ST28).

Adder 56 is when receiving coding prediction error signal from inverse quantization inverse transformation portion 55, logical Cross this coding prediction error signal and represent by infra-frame prediction portion 53 or motion compensated prediction portion The prediction signal of the 54 prognostic chart pictures generated is added and generates decoded picture, would indicate that this decoded picture Decoded picture signal be stored in infra-frame prediction memorizer 57, and by defeated for this decoded picture signal Go out to loop filter portion 58 (step ST29).

The process of step ST23～ST29 is until the whole encoding block B for layering segmentationⁿPlace Reason terminates to be repeated (step ST30) before.

Loop filter portion 58 is when receiving decoded picture signal from adder 56, and compensation comprises In the coding distortion of this decoded picture signal, the decoded picture signal after being compensated by coding distortion represents Decoded picture as being stored in motion-compensated predicted frames memorizer 59 (step with reference to image ST31)。

The Filtering Processing carried out by loop filter portion 58 both can be with having translated of exporting from adder 56 Maximum encoding block or the respective encoding block of code picture signal are that unit is carried out, it is also possible to defeated Collect 1 picture amount after having gone out the decoded picture signal suitable with the macro block of 1 picture amount to carry out.

It will be apparent from the above that, according to present embodiment 1, owing to being configured to dynamic image encoding device Infra-frame prediction portion 4 uses the encoded picture signal in frame, by implementing intra-prediction process Come in delta frame prognostic chart as time, from the wave filter of pre-prepd more than 1, according to filtering The state of the various parameters involved by coding processing object block selects wave filter, uses this filtering Prognostic chart picture is implemented Filtering Processing by device, so playing the prediction that can reduce occurring locally Error, the effect of raising image quality.

It addition, according to present embodiment 1, because being configured to infra-frame prediction portion 4 to consider (1)～(4) The parameter of more than at least 1 in parameter selects wave filter, it is possible to is inhibited and is carrying out The edge of encoded object image non-linearly distortion a littlely or because of angular deviation during directional prediction And the point prediction error produced and when being averaged value prediction with adjacent encoded letter Number Loss of continuity and the effect of the forecast error of the boundary member of block that produces, play and can change The effect of kind predictive efficiency.Described (1) is subregion P_i ⁿSize (l_i ⁿ×m_i ⁿ), (2) be bag Be contained in the quantization parameter of coded prediction error parameter, (3) are to make when the generation of medium range forecast image Encoded pixel group and the distance of Filtering Processing subject pixels, (4) be generate medium range forecast figure The index value of intra prediction mode during picture.

According to present embodiment 1, owing to being configured to the infra-frame prediction portion 53 of dynamic image decoding device Use the picture signal decoded in frame, by pre-in implementing intra-prediction process to carry out delta frame During altimetric image, from the wave filter of pre-prepd more than 1, according to Filtering Processing object block The state of the various parameters involved by decoding selects wave filter, uses this wave filter to prognostic chart picture Implementing Filtering Processing, can reduce, so playing, the forecast error occurred locally, at motion diagram Also the intra-prediction image generating as code translator side and generating in dynamic image encoding device side The effect of the same intra-prediction image.

It addition, according to present embodiment 1, because being configured to infra-frame prediction portion 53 to consider (1)～(4) The parameter of more than at least 1 in parameter selects wave filter, it is possible to is inhibited and is carrying out The edge of encoded object image non-linearly distortion a littlely or because of angular deviation during directional prediction And the point prediction error produced and when being averaged value prediction with adjacent encoded letter Number Loss of continuity and the effect of the forecast error of the boundary member of block that produces, play in motion The infra-frame prediction figure that image decoder side also generates and generates in dynamic image encoding device side The effect of the intra-prediction image as.Described (1) is subregion P_i ⁿSize (l_i ⁿ×m_i ⁿ)、 (2) be contained within the quantization parameter of coded prediction error parameter, (3) are at medium range forecast image Generation time use encoded pixel group and the distance of Filtering Processing subject pixels, (4) be generate The index value of intra prediction mode during medium range forecast image.

Embodiment 2

In above-mentioned embodiment 1, representing following such mode, i.e., infra-frame prediction portion 4 makes By the encoded picture signal in frame, by predicting in implementing intra-prediction process to carry out delta frame During image, from the wave filter of pre-prepd more than 1, according to the volume of Filtering Processing object block The state of the various parameters involved by Ma selects wave filter, uses this wave filter real to prognostic chart picture Execute Filtering Processing, but when devising the square error between the block of coded object and prognostic chart picture and being Little Wiener filter, uses above-mentioned Wiener filter and uses from pre-prepd more than 1 The wave filter selected in wave filter is compared, it was predicted that in the case of the minimizing degree of error uprises, it is possible to To replace its wave filter selected, use above-mentioned Wiener filter, prognostic chart picture is implemented at filtering Reason.

Hereinafter, process content is specifically described.

In above-mentioned embodiment 1, infra-frame prediction portion 4,53 is from pre-prepd more than 1 In wave filter, select according to the state of the various parameters involved by the coding of Filtering Processing object block Wave filter.But, in the case of considering that 4 parameters (1)～(4) select wave filter, can To select suitable wave filter from selection candidate, and when there is optimum filtering beyond selection candidate In the case of device, it is impossible to carry out " optimal Filtering Processing ".

In present embodiment 2, it is characterised in that: fill at moving image encoding in units of picture Put side design optimum filter and implement Filtering Processing, and the filter coefficient etc. to this wave filter Encode, in dynamic image decoding device side by decoding this filter coefficient etc., implement to make By the Filtering Processing of this wave filter.

The infra-frame prediction portion 4 of dynamic image encoding device is the same with above-mentioned embodiment 1, by right Encoding block BⁿSubregion PⁿImplement intra-prediction process carry out delta frame in prognostic chart as P_i ⁿ。

It addition, infra-frame prediction portion 4 is by the method the same with above-mentioned embodiment 1, from preparing in advance The wave filter of more than 1 in, according to the various parameters involved by the coding of Filtering Processing object block State select wave filter, use this wave filter to intra-prediction image P_i ⁿIt is filtered processing.

The infra-frame prediction portion 4 whole encoding block B in pictureⁿIn determine infra-frame prediction parameter After, to the region (there is the region of same filter index) using same wave filter in picture This region of each design in input picture and the square error of intra-prediction image and (target area Mean square error in territory) it is minimum Wiener filter.

Wiener filter can utilize medium range forecast picture signal s ' autocorrelation matrix R_s’s’, input Picture signal s and medium range forecast picture signal s ' cross-correlation matrix R_ss’According to below formula (4) Obtain coefficients w.Matrix R_s’s’And R_ss’Size corresponding with required filter tap number.

w = R_{s^{'} s^{'}}^{- 1} \cdot R_{{ss}^{'}} - - - (4)

Infra-frame prediction portion 4 is when designing Wiener filter, if using this Wiener filter to implement filtering The square error in wave filter design object region during process and be D1, if coding relates to this wiener Encoding amount during information (such as, filter coefficient) of wave filter is R1, if using with above-mentioned The wave filter in the case of Filtering Processing implemented by the wave filter of the method choice that embodiment 1 is same Square error in design object region and be D2, confirms whether following formula (5) is set up.

D1+ λ R1 ＜ D2 (5)

Wherein, λ is constant.

Infra-frame prediction portion 4, in the case of formula (5) is set up, replaces with same with above-mentioned embodiment 1 The method of sample carries out the wave filter selected, and uses this Wiener filter to implement Filtering Processing.

On the other hand, in the case of formula (5) is invalid, use with same with above-mentioned embodiment 1 The wave filter that the method for sample carries out selecting is to implement Filtering Processing.

Here, be evaluated with square error and D1, D2, but it is not limited to this, replacement square Error and D1, D2, it is possible to use the absolute value of error and other the expression predicted distortion waited Yardstick is evaluated.

Infra-frame prediction portion 4, in the case of using Wiener filter to implement Filtering Processing, needs to represent The filter coefficient of this Wiener filter, the wave filter which indexes is replaced into Wiener filter Filter update information.

Specifically, when the Filtering Processing by employing wave filter Selection parameter can be selected The quantity of wave filter is set to L, in the case of the index to each wave filter distribution 0～L-1, for respectively Individual index, needs in the case of using the Wiener filter that designs by " 1 " value as wave filter more Fresh information encodes, by " 0 " value as wave filter in the case of using pre-prepd wave filter More fresh information encodes.

The filter update information exported from infra-frame prediction portion 4 is carried out by variable length code portion 13 can Become length coding, the in the bitstream coded data of this filter update information of multiplexing.

Here, though it is shown that to using each region of identical wave filter in picture, design Input picture in this region and the mean square error of prognostic chart picture are the feelings of minimum Wiener filter Condition but it also may be configured to other the specific territorial units not being picture unit, for use The mean square of the input picture in this region and prognostic chart picture is designed in each region of identical wave filter Error is minimum Wiener filter, it is also possible to be only a certain specific picture to be filtered device to set Meter, or only (such as, additional Scene change detection merit in the case of corresponding with specific condition Can, in the case of the picture of scene changes being detected), carry out the design of above-mentioned wave filter.

The variable-length decoder unit 51 of dynamic image decoding device is from the coded number being multiplexed to bit stream According to, filter update information is carried out variable-length decoding.

Infra-frame prediction portion 53 is in the same manner as above-mentioned embodiment 1, by encoding block BⁿSubregion P_i ⁿImplement intra-prediction process carry out delta frame in prognostic chart as P_i ⁿ。

Infra-frame prediction portion 53 when receiving filter update information from variable-length decoder unit 51, With reference to this filter update information, confirm in the wave filter of corresponding index with or without renewal.

Infra-frame prediction portion 53 is when some region of wave filter is replaced into wiener filter in the result confirmed In the case of ripple device, read the wave filter system of the Wiener filter being contained in this filter update information Number, determines this Wiener filter, uses this Wiener filter to implement intra-prediction image P_i ⁿFilter Ripple processes.

On the other hand, in the region not being replaced into Wiener filter, with above-mentioned embodiment 1 The same method selects wave filter, uses this wave filter to implement intra-prediction image P_i ⁿFiltering Process.

It will be apparent from the above that, according to present embodiment 2, because being configured to the block designing coded object with pre- Square error between altimetric image and be minimum Wiener filter, when using this Wiener filter and making Compare with the wave filter selected from the wave filter of pre-prepd more than 1, it was predicted that subtracting of error In the case of few degree uprises, its wave filter selected is replaced to use this Wiener filter, in advance Altimetric image implements Filtering Processing, reduces further compared with above-mentioned embodiment 1 it is possible to play The effect of the forecast error occurred locally.

And, the freedom of each embodiment can be carried out in the range of its invention in the present application Combination or each embodiment arbitrary element deformation or at each embodiment The arbitrary element of middle omission.

Industrial applicability

The present invention is suitable to need the picture coding device of coded image expeditiously, it addition, be suitable to need Will be expeditiously to being coded of the image decoder that image decodes.

Claims

1. a picture coding device, it is characterised in that possess:

Intraprediction unit, at the coding mode that the encoding block obtained with splitting input picture is corresponding In the case of being intra-frame encoding mode, real to the block of the unit that the prediction becoming described encoding block processes Execute intra-prediction process, generate prognostic chart picture,

In described intraprediction unit, in the case of the value prediction that is averaged, using conduct The centre becoming the meansigma methods of multiple adjacent pixels of the block of the unit that described prediction processes and be prescribed Predictive value and the adjacent pixels of prediction subject pixels, the list that the prediction becoming described encoding block is processed When the described prediction subject pixels being positioned at upper end and left end in the block of position implements Filtering Processing,

In the block of unit that described prediction subject pixels is the prediction process becoming described encoding block In the case of the pixel of the left end being positioned at described piece beyond the pixel of upper left, by pre-for described centre Coefficient involved by measured value is set to 3/4, by the pixel institute adjacent with the left side of described prediction subject pixels The coefficient related to is set to 1/4.

2. an image decoder, it is characterised in that possess:

Intraprediction unit, is the situation of intra-frame encoding mode at the coding mode involved by encoding block Under, the block of the unit that the prediction becoming described encoding block processes is implemented intra-prediction process, generates Prognostic chart picture,

In described intraprediction unit, in the case of infra-frame prediction parameter represents meansigma methods prediction, The quilt in the meansigma methods of multiple adjacent pixels of the block used as the unit becoming described prediction process The intermediate predictor of regulation and the adjacent pixels of prediction subject pixels, to becoming the pre-of described encoding block The described prediction subject pixels being positioned at upper end and left end in the block of the unit that survey processes is implemented at filtering During reason,

3. a method for encoding images, it is characterised in that possess:

Intra-prediction process step, at the coding that the encoding block obtained with splitting input picture is corresponding In the case of pattern is intra-frame encoding mode, to the unit that the prediction becoming described encoding block processes Block implements intra-prediction process, generates prognostic chart picture,

In described intra-prediction process step, in the case of the value prediction that is averaged, using It is prescribed as the meansigma methods of multiple adjacent pixels of block becoming the unit that described prediction processes Intermediate predictor and the adjacent pixels of prediction subject pixels, the prediction to becoming described encoding block processes Unit block in be positioned at upper end and left end described prediction subject pixels implement Filtering Processing time,

4. an image decoding method, it is characterised in that possess:

Intra-prediction process step, is intra-frame encoding mode at the coding mode involved by encoding block In the case of, the block of the unit that the prediction becoming described encoding block processes is implemented intra-prediction process, Generate prognostic chart picture,

In described intra-prediction process step, represent, in infra-frame prediction parameter, the feelings that meansigma methods is predicted Under condition, using the average of the multiple adjacent pixels as the block becoming the unit that described prediction processes Value and the intermediate predictor that is prescribed and the adjacent pixels of prediction subject pixels, to becoming described coding The described prediction subject pixels being positioned at upper end and left end in the block of the unit that the prediction of block processes is implemented During Filtering Processing,